Résumé : A cubic tree is a tree with (Formula presented.) leaves in which every internal vertex has degree exactly 3. Any such tree (Formula presented.) can be encoded by a Path-Length Matrix (PLM), that is, an (Formula presented.) integer matrix whose (Formula presented.) th entry gives the number of edges in the unique path between leaves (Formula presented.) and (Formula presented.) in (Formula presented.). The convex hull of all PLMs associated with cubic trees on (Formula presented.) leaves defines the PLM-polytope. In this article, we prove that optimizing a linear function over the PLM-polytope is strongly (Formula presented.) -hard. This result is motivated by the balanced minimum evolution problem, a highly nonlinear (Formula presented.) -hard network design problem arising in computational biology, whose underlying combinatorial structure is captured by the PLM-polytope.